Deliberate enhancement of rainfall using desert plantations

Significance Our desert plantation concept aligns closely with research into biological carbon sequestration solutions but uniquely extends into the purview of deliberate rainfall enhancement. With this synergy of carbon sequestration and regional weather modification, we can counteract water scarcity and desertification while minimizing conflicts with food croplands. We have demonstrated that large plantations do enhance rainfall in arid regions and identified the underlying process chain. By using this knowledge we have developed a global index to assess which deserts are most favorable for weather modification and discuss how rainfall impacts can be intensified using agricultural methods. This potential for rainfall enhancement and carbon sequestration makes the research extremely interesting for the scientific community and for society.


5
We emphasize the importance of using well-tested physics scheme set within WRF, to correctly simulate the process chain operating from the land surface to the troposphere including cloud microphysics. Important studies demonstrating the performance on this study's parameterizations include the YSU boundary layer scheme 10,11,12,13 , the Noah LSM 4,14,15,16 and the Morrison microphysics scheme 17,18,19,20 .
The scenarios were designed to demonstrate the impact of a 100 × 100 km rectangular jojoba plantation (Simmondsia chinensis) on local weather within two different regions: the arid region of Oman and the arid/semi-arid region of Israel (domains and plantations shown in Fig. S1). Database (HWSD) was included to optimize soil texture mapping 23 , and plant and soil parameters were selected based on a one year jojoba simulation, and validation with in-situ measurements of jojoba 24 . Important plant parameters were leaf stomatal resistance, albedo, canopy geometry/roughness, and soil properties such as saturated hydraulic conductivity and porosityobtained from local surveys and literature (See Table S4).

Statistical analysis
Analysis was carried out on the two impact scenarios, by grouping the regional data together and splitting it into CI and non-CI days (See Fig. S2). CTP is calculated as: where is temperature and is dewpoint temperature [˚K].
Wind shear was calculated as the difference in wind speed between 850 and 700 hPa [ms -1 hPa -1 ]. All variables were calculated as an average between 07:00 and 09:00 AM local time to characterize the morning pre-convective environment.
Two-sample t-tests were then conducted in Originlab, with a Welch correction to account for nonnormal distributions and statistical significance indicated at p<0.05. Significant differences in means were observed for all 4 variables (Table S5 and Fig. S2).

Derivation of Global Feedback Index (GFI)
The intention was to assess different arid regions for their climate suitability for weather modification by identifying the ideal conditions for CI, derived from the WRF IMPACT simulations, and applying them to analyses of global climate data. In this way we could identify ideal climate conditions over the Earth's arid regions. For this purpose, we used nine years of monthly means of daily means from the ECMWF ERA5 reanalysis global dataset (2009-2017) 29 .
The first task was to identify the ideal conditions for CI, based on the WRF Impact simulations.
From Fig. S2, we observed that the mean values of CTP, HIlow and Shear on CI days were greater than one standard deviation (1σ) away from the non-CI means ( occur, it is more likely to be triggered by larger scale conditions, e.g. when embedded in frontal systems 22 . It is assumed that all quantities need to be optimal simultaneously for CI to occur, because a sub-optimal value for any one variable may disrupt the process chain altogether. To identify the optimal variable-space for CI, the statistical analyses seen in Table S5 and Fig. S2 were used as a basis. The non-CI day means were statistically different from the CI days (>1σ), and we therefore considered variables as optimal within 1σ around the CI mean. Even then, to be sure of a 'conservative' variable-space we selected even tighter thresholds. Figure S3 shows our score algorithm used for identifying optimal and sub-optimal variable-spaces. 12 For CTP we selected values above 0.5σ below the CI mean as being semi-optimal (score 0.5), and values above the CI mean as optimal (score 1), and values below 0.5σ below the mean as suboptimal (score 0). For HIlow we assigned scores of 0.5 and 1 to ranges 1σ above and below the mean and 0.5σ above and below the mean, respectively. Values outside 1σ above and below the mean are suboptimal (score 0).
From the ERA5 dataset, we calculated CTP, HIlow and generated mean statistics for each month of the year, over the 2009-2017 period and also for the year of simulation, 2012 (Figs. S4-S15). This comparison was made to assess whether the 2012 conditions were representative for the nine year period.

Highlight -Land surface-atmosphere feedbacks
Here we demonstrate the plantation impact process chain in more detail, from the land surface up to the upper troposphere. Fig. S16 shows the plantation effect close at the land surface upon heating and pressure/wind deformation, on 30 th June 2012 when a CI event with rainfall was observed (up to 9 mm day -1 in some grid points around the plantation). shown and on the right, the IMPACT run.

27
There was a clear heat-low air pressure reduction of 1 to 1.5 hPa, and this pressure gradient force together with surface friction led to a convergence of winds toward the leeside of the plantation. .00 UTC (11.00-13.00 LT), as observed in WRF-NOAH Impact run. 28 We observed clearly that the wind convergence and increased turbulence over the plantation created a strong upward mixing of heat and vapor to altitudes of over 5000 m, higher than the level of free convection. This process of de-stratifying the PBL occurs over the 3 hours between 11.00 and 13.00 local time. See also Movie S1 for a 3D animation of PBL evolution and CI on June 30 th 2012.